Wetlands Ecology and Management (2005) 13:671684 DOI 10.

1007/s11273-005-0965-9

Springer 2005

-1

A tropical freshwater wetland: II. Production, decomposition, and peat formation

Rodney A. Chimner1,2,* and Katherine C. Ewel1
USDA Forest Service, Pacic Southwest Research Station, Institute of Pacic Islands Forestry, 1151 Punchbowl Street, Ste. 323, Honolulu, HI 96813, USA; 2Current address: Natural Resource Ecology Laboratory, Colorado State University, 200 W. Lake Street, Ft. Collins, CO 80523-1499, USA; *Author for correspondence (e-mail: rchimner@nrel.colostate.edu; phone: +1-970-491-2162; fax: +1-970-491-1965)
Received 23 December 2004; accepted in revised form 19 January 2005
1

Key words: Carbon cycling, Decomposition, Kosrae, Micronesia, Peatlands, Plant production

Abstract As much as 10% of the total carbon stored in peatlands occurs in the tropics. Although tropical peatlands are poorly understood scientically, they are increasingly exploited for a variety of human uses. Our objective was to measure baseline carbon cycling data in one type of tropical peatland in order to understand better how peat accumulates in these ecosystems. Average plant production for two study sites on the island of Kosrae in the Federated States of Micronesia over 2 year was 1122 g C m2 year1, of which 1058 g C m2 year1 was aboveground plant production (bole, buttress and litterfall). Although leaves contributed a high proportion of total plant productivity, their rapid decomposition left little carbon for peat accumulation. In contrast, ne roots only contributed $10% to plant productivity, but their slow decomposition allowed them to accumulate as peat. Wood (branches and stems) probably contributed the most carbon to the formation of peat. Despite being on the soil surface, small branches decomposed more slowly than leaves because of their high C:N and lignin:N ratios. In summary, we suggest that tropical peatlands in Micronesia accumulate peat not because of high plant production but rather because of slow decomposition of roots and wood under anaerobic conditions that result from nearly constant high water levels.

Introduction Tropical peatlands represent as much as 10% of the total carbon stored in peatlands (Immirzi et al. 1992), which cover roughly 4 million km2 or 3% of the Earths land surface (Maltby and Proctor 1996). They can be up to 10 m deep, and accumulation rates can be rapid, averaging 45 mm/ year, with some sites as fast as 510 mm/year (Maas 1996). These rates are signicantly faster than in most temperate and boreal peatlands,

which accumulate less than 1 mm/year, often slower than 0.5 mm/year (Gorham 1991; Gorham et al. 2003). Although tropical peatlands are poorly understood scientically, they are increasingly exploited for peat extraction, sheries, fossil fuels, minerals, forestry, and agriculture (De la Cruz 1986; Rieley et al. 1996; Junk 2002). Tropical peatlands are at greater risk than remote temperate and boreal peatlands (such as in Canada and Russia), because they are often situated in densely populated low-lying coastal areas (Rieley et al. 1996). With

672 high population growth in many tropical countries, tropical peatlands are facing increasing rates of alteration and destruction (Vijarnsorn 1996). In order to understand and manage peatlands it is imperative to understand the processes that regulate carbon cycling and the rate of peat accumulation. Peatlands accumulate carbon because plant production is faster than decomposition, allowing the build up of peat, which is partially decomposed plant material. Peat accumulates in subarctic and boreal peatlands because of low decomposition rates rather than high net annual primary production (Malmer 1986; Francez and Vasander 1995). However, tropical peatlands occur in consistently hot and often humid conditions, which are normally associated with both high production rates and rapid decomposition rates. Several explanations are possible for how peat accumulates in the tropics. Tropical peatlands may function similarly to peatlands in other regions because higher temperatures in the tropics increase both plant production and decomposition rates similarly. Alternatively, hot and humid tropical conditions might cause key changes to the carbon cycling in peatlands. Aboveground plant production is believed to be the primary source of peat (Clymo 1983). This paradigm was developed for Sphagnum moss peatlands, which is the dominant peat-producing plant in many regions and has no roots. However, tropical peatlands are predominantly forested with no moss cover (Rieley and Ahmad-Shah 1996). The high temperatures and more aerobic conditions on the surface of tropical peatlands may speed up the decomposition of leaves and wood, suggesting that roots could be more important to peat accumulation in the tropics. Although roots only contribute 1030% of the total plant biomass in northern forested wetlands (Trettin et al. 1995), they occur below the soil surface in almost continually anaerobic conditions. The few documented root decomposition studies have shown that roots normally decompose slower than aboveground components (e.g.. Day et al. 1989; Thormann et al. 2001). Recent models also predict that peat could be composed of predominantly root material, with leaves contributing little to carbon storage (Chen and Twilley 1999; Chimner et al. 2002). The future is uncertain for tropical peatlands, and the rate of disturbance and destruction is very high, yet there is little scientic information to help with management, sustainable use, or protection. Understanding how peat accumulates and how carbon cycles in the tropics will not only provide information for management of tropical peatlands, but it also will help in elucidating how carbon accumulates in all peatland types. The objective of this study was to obtain baseline carbon cycling data for one type of tropical forested peatland in order to understand better how peat accumulates in these wetlands.

Study sites This study was conducted on the island of Kosrae (109 km2), a high volcanic island in the Federated States of Micronesia in the western Pacic Ocean (Figure 1: 519 N, 16300 E). Kosrae is 1.4 million years old and has a rugged, forested interior, a narrow coastal plain lined by mangroves around most of the island, and a fringing reef. Rainfall and temperature are aseasonal. Average daily temperatures during our study (May 2001May 2003) ranged between 25 and 30 C, with an average of 27 C. Total rainfall during our study was 5900 mm/year during May 2001May 2002, and 4900 mm/year during May 2002May 2003. The long-term mean (19542001) for Kosrae is 5050 mm/year (NOAA Cooperative Weather Station 914395). September 2001 was the driest month with 250 mm, whereas December 2001 was the wettest with 1080 mm. Peatlands are not usually associated with tropical islands. However, freshwater peatlands are estimated to cover 807 ha, or 1.2% of total land area, in the Federated States of Micronesia (USDA Soil Surveys 1983). Mangrove peatlands represent another 2446 ha, or 3.7% of total land area. Freshwater peatlands are often located directly behind mangrove forests or beach strands (Merlin et al. 1993). In Kosrae, they are dominated by Terminalia carolinensis Kanehira (hereafter called Terminalia), which is a large tree (up to 35 m tall) with a distinctive pagoda-shaped crown and large buttresses (Maxwell 1982). Endemic to Micronesia, this species is common only on the island of Kosrae, where it occupies about 300 ha (Miller 1984). On the neighboring island of

673

Figure 1. Location of the island of Kosrae, Federated States of Micronesia, in the western Pacic Ocean. Study sites are located in the Yewak region on the east side and the Yela region on the northwest side of Kosrae.

Pohnpei, it is much less common due to widespread conversion of wetlands to agriculture (Merlin et al. 1992). We established two study sites. The rst was located in the Yela basin (519.560 N, 16256.88 E) in the municipality of Tafunsak on the northwest side of Kosrae (Figure 1), in one of the largest (68 ha) and most pristine forested wetlands in Micronesia. There are currently no roads into the area, although a road across the basin or around the rim has been proposed. The second study site was located in an 18-ha forested peatland in the Yewak region (517.615 N, 16301.564 E) in the municipality of Malem on the east side of Kosrae (Figure 1). Aerial photos from 1944 indicate that the western side of the peatland was intensively farmed until the end of

World War II. Subsequent aerial photos show that part of the agricultural portion of the peatland was converted to agroforestry and part was left to regenerate naturally. The eastern side of the peatland was not farmed intensively, but many large Terminalia trees have been harvested from it over the past 50 years.

Methods Stand dynamics and plant production Circular 0.04-ha plots were established in the center of each study site. Within each plot, each tree >2.5 cm in diameter was identied, and its diameter was measured at 50 or 130 cm above the

674 highest buttress. Height of each tree was measured using a relaskop. Tree bands were installed (at the same height where diameter was measured) on ve trees at each site: at Yela we used tree band data from ve of the Terminalia trees being measured as part of another study (Allen et al. unpublished MS); at Yewak, we installed bands on ve Terminalia trees and ve trees of another common species, Horseldianunu Kanehira. To translate diameter growth into wood production, we used previously developed outsidebark stem volume equations for a third tree species, Barringtonia racemosa (L.) Spreng. (Ewel et al. 2003), but no stem volume equations were available for Terminalia or H. nunu. We therefore derived regression equations to estimate outsidebark total tree volume (m3) from stand inventory data collected in Kosrae (MacLean et al. 1988; Ewel et al. 2003). Volume equations were derived by regressing volume against diameter. The nal equations used to calculate stem volume are: Terminalia m3 =tree) 0:25 expdbh 0:044474 : R2 0:66 H: nunu m3 =tree 0:000128 dbh2:363 : R2 0:89 B: racemosa m3 =tree 0:03134 0:000446 dbh2 : R2 0:98 Ewel et al. 2003 Terminalia trees have large buttresses that were not measured in the stand inventory and are consequently not incorporated into the stem volume equations. To correct for this, we developed separate buttress volume equations. We measured the height, width, and length of all buttress segments on seven trees from both sites that ranged in size from 29.3 to 84.5 cm dbh. The volume of each buttress segment was calculated using the formula for a right triangle. Total buttress volume for each tree was calculated as the sum of all buttress segment volumes. Buttress volume was signicantly correlated to dbh and was described by an exponential growth function: Buttress Volume (m3 /tree) 0:076 expdbh 0:038 : R2 0:99 This equation was used to calculate buttress volume on all Terminalia trees. Initial wood biomass was calculated using the allometric equations described above. Final wood biomass was calculated by using the new dbh after 1 year as determined for each tree species from the tree band data and recalculating volume using the allometric equations. Wood production was calculated by subtracting the nal wood biomass from the initial wood biomass. Litterfall was captured at each site in eight 1.0 m 1.0 m litter traps constructed by attaching shade cloth to four 2.5 cm PVC pipes inserted into the soil. Litter was collected at monthly intervals for 2 year, sorted into components (leaves, reproductive parts, ne branches (<10 cm), and miscellaneous), oven-dried for 3 days, and weighed. To estimate large-branch fall (>10 cm) we measured branches that fell into three replicate 5 5 m plots at each site. After all branches had been removed from the plots, we collected newly fallen branches at monthly intervals. They were dried and weighed. Belowground ne root growth was calculated using in-growth root bags (Neill 1992; Weltzin et al. 2000; Chimner and Ewel 2004). Mesh bags 30 cm long and 10.1 cm in diameter made from berglass window screen (1.5 mm mesh) were lled with native root-free peat and inserted in an augured hole to a depth of 30 cm. Eight replicates were installed in each site. Roots were collected from the mesh bags, washed free of sediment, oven dried, weighed, and analyzed for carbon content using a CHN analyzer. Decomposition bags Decomposition bags were made for leaves, ne branches (<10 mm), and ne roots (<2 mm). Leaves were decomposed in 20 cm20 cm mesh bags constructed from 1.5 mm berglass screen. Each mesh bag was lled with 6 g of air-dried leaves collected from freshly fallen litter. Fine branches were collected loose on the forest oor or from downed trees. After air-drying, 6 g were decomposed in 10 cm20 cm mesh bags. Live ne roots were collected in the eld, brought back to the lab, washed, and air-dried. Three grams of airdried roots were inserted into 10 cm10 cm 1.5 mm mesh berglass bags. Air-dried to

675 oven-dried conversion factors for leaves, ne branches, and ne roots were calculated from subsamples. Decomposition bags for leaves and stems were placed on the soil surface, while root bags were buried at a depth of 5 cm. Five bags of each type were collected at 2, 4, 8, 12, 17, and 36 week. Upon collection, decomposition bags were stored at 4 C in labeled Ziploc bags until processed. Each decomposition bag was washed in water to remove sediments, dried at 70 C for 3 days, weighed, and ground for further chemical analysis. The percent mass remaining for each plant type and site was calculated using the following equation: MR Xo X=Xo 100; where Xo represents initial dry mass and X = mass remaining. First-order and second-order exponential decay models were tted to the decomposition data for each plant type and site using SYSTAT 10.0 non-linear regression (SPSS 2000). The rst-order model was MR=100*ekt, where MR = mass remaining at time t in years, and k was the decay constant (Olson 1963; Wieder and Lang 1982). The second-order model was MR=S * ekt+(1S)* eht, where MR = mass remaining at time t in years, S was the initial proportion of labile material, (1S) is the initial proportion of resistant material, k was the rate constant for the labile material, and h was the rate constant for the resistant material (Hunt 1977; Wieder and Lang 1982). 2000, LECO Corporation, St. Joseph, Michigan, USA). Carbon isotope measurements were conducted on a Caria Erba NC 2500 Elemental Analyzer with conow II interface coupled to a Finnigan Delta-S mass spectrometer with a source upgrade.

Ancillary measurements Groundwater levels were measured in monitoring wells constructed from 3.8-cm diameter slotted PVC pipes in each peatland. Wells were installed with a standard bucket auger of the same diameter as the well casing. Water tables were measured four times during the year. Water table levels were also continuously measured in Yewak with a YSI 600 XLM sonde (YSI, Inc., Yellow Springs, Ohio). Specic conductivity, pH, and redox potential were measured during each CO2 sampling period. Redox was measured using referenced platinum electrodes inserted into the soil to a depth of 5 cm. Specic conductivity and pH were measured in the surface water using a portable meter.

Results Environmental and soil parameters There were few dierences in pH, specic conductivity, or redox levels between sites (Table 1). Soil temperature varied only slightly during the study period, averaging 25 C. Water table levels were maintained near the soil surface during the study period with uctuations no greater than 10 cm. Bulk density was also similar between the sites, averaging 0.1 g/cm3 (Table 1). Dierences in peat chemistry that may reect hydrogeomorphic dierences between the sites were detected, however. The Yela site is located within a broad river valley that is ooded regularly during high rainfall events and consequently has a high amount of sediment deposited in the peatland. The peat there had an organic carbon content of only 20% (Table 2). This is in contrast to almost 50% organic carbon content for the peat at Yewak, which is not located in a river valley and has little sediment deposition. Nitrogen and lignin concentrations were also signicantly lower in Yela. The C:N ratio was signicantly lower in Yela, but lignin:N ratios were similar.

Substrate quality Five replicate peat samples (05 cm) were collected from each site and analyzed for lignin, cellulose, and carbon and nitrogen contents. Plant material from the decomposition bags were also analyzed for lignin, cellulose, carbon, nitrogen, and d13C contents at every retrieval period. The ve replicate decomposition bags at each site were pooled for analysis. Peat and plant samples were analyzed at the Agricultural Diagnostic Service Center, Sherman Laboratory, University of Hawaii. Lignin and cellulose concentrations were determined using the procedure of Goehring and Van Soest (1970). Total carbon and nitrogen were determined using dry combustion (LECO CHN-

676
Table 1. Mean values of environmental parameters of the study sites. Yela PH Redox (mv) Specic conductivity (lS) Soil temperature (C) Water table (cm) Bulk density (g/cm3) 6.4 133 119 24.8 0.1 0.11(0.01) Yewak 5.9 154 133 25.6 3.4 0.13 (0.03)

Surface water was used for determining pH and specic conductivity. Redox and soil temperature were measured at depths of 5 cm. Bulk densities are an average of ve samples, whereas other values are averages of four measurements.

Forest structure and production Three tree species were encountered in our study sites (Table 3). Terminalia and H. nunu were found at both sites, whereas B. racemosa was only encountered at Yela. The Terminalia in Yela were fewer in number but had twice the height, diameter, and basal area than in Yewak. H. nunu had similar heights and basal area at both sites but were much larger and fewer in number at Yela.

Many small B. racemosa were found in the understory at Yela. Total wood volume was twice as high in Yela with 666 vs. 332 m3/ha in Yewak (Table 3). The buttress volume of Terminalia averaged 88 and 49 m3/ha, which accounted for 16 and 20% of the total stem volume in Yela and Yewak, respectively. Increment growth for Terminalia averaged 3.8 and 12.4 mm/year in Yela and Yewak, respectively. Despite having half the wood volume, Yewak had greater wood production due to higher increment growth, averaging 861 g m2 year1, compared to 658 g m2 year1 for Yela. Annual litterfall did not dier signicantly (p>0.10) between sites during either year (Figure 2). Annual litterfall rates for Yewak were not signicantly dierent between years (p = 0.4); they totaled 1.33 and 1.65 kg m2 year1 for 2 years, with an average of 124 g m2 month1. However, annual litterfall rates for Yela were signicantly higher in the second year (2.04 kg m2 year1) than the rst (0.98 kg m2 year1) (p = 0.003) for an average litterfall of 126 g m2 month1 for both years. Time of the year was a signicant factor for litterfall (p<0.001) with the highest rates

Table 2. Mean, standard deviation (s.d.), and p-values from t-tests of peat chemistry from Yela and Yewak. Yela Mean Carbon (%) Nitrogen (%) Lignin (%) C:N Lignin:N 19.9 1.20 23.5 16.6 19.5 s.d. 2.62 0.17 4.41 1.44 2.14 Yewak Mean 47.0 2.34 47.9 20.3 20.56 s.d. 0.99 0.23 3.54 2.54 1.84 <0.001 <0.001 <0.001 0.029 0.44 p

Table 3. Stand characteristic, above ground volume of stems and buttresses, and above ground growth of trees in Yela and Yewak. Yela Terminalia carolinensis Mean height (m) Mean dbh (cm) Number of trees (ha) Basal area (m2/ha) Butters volume (m3/ha) Stem volume(m3/ha) Total volume (m3/ha) Wood productiona (kg C ha2 year1)
a

Combined Barringtonia racemosa 11 17 400 9 38 38 490 Horseldia nunu 18 54 50 13 92 92 326 (16) (28) 550 51 88 578 666 2830

Yewak Terminalia carolinensis 17 23 250 12 49 187 235 2907 Horseldia nunu 15 15 200 16 97 97 795

Combined

32 59 100 29 88 448 536 2014

(16) (26) 450 28 49 284 332 3702

Assuming carbon content of 43% and wood density of 522 kg/m3 (T. catappa).

677

Figure 2. Litterfall patterns at Yewak and Yela sites for 2 year.

occurring during the winter months for both sites. For the two sites combined, litterfall consisted of 63% leaves, 21% reproductive material, 9% branches, and 7% miscellaneous material. Fine root biomass did not dier signicantly (p = 0.08) between Yela and Yewak, averaging 272 g/m2. An average of 80% of the root biomass was found in the top 15 cm of the cores. Fine root production did not dier between the two sites (p = 0.15), averaging 300 g m2 year1. Plant production for this study was calculated as the sum of ne root production, litterfall,

branchfall, and wood production. We do not have data for coarse root production, but this is usually a small component of plant production. Annual plant production for Yewak totaled 1037 and 1175 g C m2 year1 for 2 years, averaging 1106 g C m2 year1 (Figure 3). Annual plant production for Yela totaled 864 and 1409 g C m2 year1 for 2 years, averaging 1137 g C m2 year1. Overall, litter accounted for 36% and total wood production, including branches, accounted for roughly 30% and ne root production accounted for 12% of total production.

678

Figure 3. Mass remaining for leaves (solid line) and ne root (dashed line) decomposition bags over a 36-week period. Regression lines are the best t double exponential regressions, and error bars are standard errors.

Decomposition Decomposition rates were similar between sites with no signicant dierences for leaves, branches, or roots (p>0.05). Decomposition of leaves occurred rapidly with an average of 50% remaining after 8 week and only 20% remaining after 17 week (Figure 4). Leaves had high k-values of 3.66 and 5.24 for the single exponential model at

Yela and Yewak, respectively (Table 3). However, after 17 week decomposition slowed dramatically with no signicant dierence between mass remaining after 17 and 36 week (p>0.05). Decomposition rates were slower for both roots and branches (Figure 4). Roughly 85% of branch mass remained after 8 week, and 60% remained after 37 week. This is similar to the roots, which had roughly 81% remaining after 8 week and 68%

679

Figure 4. Tentative carbon budget for forested peatlands in Kosrae. Square boxes are pools, circles are uxes, and arrows are residual carbon after 1 year of decomposition. Pools are in g C m2 and uxes and residual carbon are in g C m2 year1. 1Data from Allen et al. (unpublished MS), 2Data from Chimner (2004).

remaining after 37 week. For the single exponential model, both branches and roots had k-values less than one. Branches had low k-values of 0.8 and 0.7 for the single exponential model; roots also had low k-values of 1.0 and 0.6 at Yela and Yewak, respectively (Table 4). The double exponential models t as well as or better than the single exponential models for all components and sites (Table 4). They predicted that the leaves had more labile material than either the branches or roots. The carbon concentrations of the dierent vegetation components were initially similar, averaging around 42% and changing little

throughout the decomposition process (Table 5). Initial nitrogen concentrations were lowest for the branches and highest for the roots. The nitrogen concentrations increased during decomposition in the leaves, but not for the ne roots and branches. Final lignin concentrations in the leaves stabilized at 812 week at about half of the initial concentration. Lignin concentrations remained the same or increased in the branches and ne roots during the decomposition process. The C:N and lignin:N ratios decreased in the leaves, but they increased or changed little in the branches and ne roots.

Table 4. Parameter estimates for single and double exponential decomposition models. Parameter Yela Leaves k R2 S k h R2 3.66 0.94 87.76 4.72 0.00 0.94 Branches 0.845 0.88 60.33 2.31 0.58 0.87 Roots 1.064 0.64 37.25 7.70 0.00 0.87 Yewak Leaves 5.24 0.94 50.45 19.3 1.63 0.98 Branches 0.70 0.87 33.74 3.91 0.00 0.98 Roots 0.58 0.85 32.70 2.94 0.00 0.88

k is the decay constant for the rst-order model. S is the initial proportion of labile material, k is the rate constant for the labile material, and h is the rate constant for the resistant material for the second-order model.

680 Discussion Mechanisms of peat formation Peatlands form when plant production exceeds decomposition, allowing organic matter to accumulate. Our plant production rates did not support the thesis that higher rates of plant productivity drive peat formation. Our average plant production for the two sites over 2 years was 1.12 kg C m2 year1, of which 1.06 kg C m 2 year1 was aboveground plant production (bole, buttress and litterfall) (Figure 5). In comparison, aboveground production rates of northern forested wetlands range from 0.100 to 1.01 kg C m2 year1 (Trettin et al. 1995). Stem production at Yewak and Yela averaged 327 g C m2 year1 compared to 48541 g C m2 year1 from the same kinds of wetlands (Lugo et al. 1988, cited in Trettin et al. 1995). Further evidence that high plant production is not important for tropical peat formation is apparent from the production rates of the dominant tree Terminalia in Kosrae. Tree growth rates of Terminalia in peat soils were signicantly slower than in non-peat soils with depth to water table the most apparent environmental factor dierentiating the two kinds of soil types (Allen et. al. unpublished MS). High water tables may have slowed not only plant production, but decomposition as well, allowing peat to form. Decomposition and soil respiration generally increase with an increase in temperature (Hogg et al. 1992; Lloyd and Taylor 1994; Raich and Potter 1995; Fang and Moncrie 2001; Silver and Miya 2001; Chimner 2004). Decomposition data for peatlands, excluding sphagnum bogs, reveal similar trends with increasing latitude (Figure 5). For example, leaves at higher latitudes lose between 20 and 40% of their mass within the rst year, whereas leaves in our tropical site lost all their mass in the rst year. Branches showed the same pattern as leaves. The temperature pattern for the limited data on ne roots was not as strong

Figure 5. Pattern of mass loss from decomposition in forested peatlands (Reader and Stewart 1972; Heal and French 1974; Bums 1976; Duever et al. 1976; Brinson 1977; Chamie and Richardson 1978; Nessel 1978; Bartsch and Moore 1985; Tupacz and Day 1990; Conner and Day 1991; Thorman and Bayley 1997; Thorman et al. 2001; Chimner and Ewel this study).

681 (Figure 5). These dierences indicate how peat can form in a tropical environment. Although leaves contributed a high proportion of total plant productivity, their rapid decomposition allowed little carbon to be available for peat accumulation. In contrast, ne roots only contributed $10% to plant productivity, but their slow decomposition allowed them to accumulate as peat (Figure 5). Although we were not able to quantify coarse root dynamics, we believe that they would also have slow decomposition rates, probably slower than the ne roots due to poorer substrate quality, and would also contribute most of their biomass to peat formation, increasing the importance of roots to peat formation even more. Wood (branches and stems) probably contributed the most carbon to the formation of peat (Figure 5). Despite being on the soil surface, small branches decomposed more slowly than leaves because of low substrate quality, as represented by their high C:N and lignin:N ratios (Table 5). If we assume that tree stems decomposed at the same rate as twigs, then stems would have added an additional 153 g C m2 year1 of carbon to the formation of peat (Figure 5). We estimated a carbon balance for Terminalia wetlands by summing residual carbon amounts left over from the decomposition of plant production (Figure 5): we calculate that these sites had 235 g C m2 year1 of residual carbon. This does not include the 1983 g C m2 year1 of miscellaneous litter fall for which we do not have decomposition rates. We assume that most of the miscellaneous litter fall was added to the peat prole as Terminalia seeds, which are large, very hard nuts that do not appear to decompose very fast. For instance, we found 380 g C m2 of Terminalia nuts in the top 30 cm of soil in Yela while measuring ne root biomass. Carbon gain can also be estimated by subtracting soil respiration from plant production (Trumbore et al. 1999; Chimner and Cooper 2003). If we assume that root respiration is half of total soil respiration, than we can estimate that these sites accumulated 407 g C m2 year1. This does not account for losses due to dissolved organic carbon or CH4 losses, which would lower the carbon gain. Combining these two indirect methods gives us an estimated accumulation of 300 g C m2 year1 for these systems. This is comparable to carbon accumulation rates in forested tropical peatlands in Thailand, which averaged 527 g C m2 year1 (Suzuki et al. 1999). Peat is often thought of as a black box with constant decay parameters or temperature-dependent rates. But organic matter is actually made up of a multitude of compounds, each with a dierent decay rate (Parton et al. 1987). The compounds can be roughly divided into three groups: non-structural carbohydrates, cellulose and

Table 5. Plant chemistry of leaves, branches, and ne roots from decomposition bags retrieved after progressive number of weeks. Yela 2 Leaves Carbon % Nitrogen % Lignin % C:N Lignin:N Branches Carbon % Nitrogen % Lignin % C:N Lignin:N Roots Carbon % Nitrogen % Lignin % C:N Lignin:N 4 8 12 17 36 Yewak 2 4 8 12 17 36

42.2 0.6 39.7 69.2 65.1 42.3 0.4 29.8 109 76 43.0 1.2 34.5 36.7 29.5

41.4 0.7 37 59.2 59.2 44.1 0.4 25.9 116 68 42.6 1.1 39.3 38.0 35.1

41.3 1.4 39.4 29.9 28.6 43.1 0.3 30.1 166 116 43.4 1.2 32 36.2 26.7

38.5 1.3 33.9 30.6 26.9 43.8 0.2 30.8 183 128 44.6 1.1 39.9 40.6 36.3

36.7 1.2 34.1 30.6 28.4 42.3 0.4 30.2 108 77 44.3 1.1 38.4 39.9 34.6

38.9 1.6 35.4 24.9 28.4 46.0 0.4 33.5 132 96 44.8 1.3 43 34.5 33.1

42.7 0.8 30.8 56.2 40.5 45.1 0.3 31.9 173 123 42.8 1.3 35.2 34.0 27.9

42.1 1.1 39.3 38.2 35.7 43.7 0.3 29 137 91 42.6 1.3 40 33.0 31.0

41.8 1.3 41.1 31.9 31.4 43.9 0.3 31.3 151 108 42.5 1.4 34 30.0 23.9

43.0 2.0 23.3 21.8 11.8 44.7 0.4 34.7 115 89 44.3 1.5 37.3 29.5 24.9

43.2 1.8 44 24.4 24.9 44.9 0.4 38.8 121 105 43.4 1.6 34.9 26.6 21.4

43.7 2.1 39.1 20.5 18.4 47.6 0.3 36.2 153 117 44.9 1.6 44.9 27.9 27.9

682 hemicellulose, and highly recalcitrant material like lignin (Verhoeven 1986). Tropical conditions should favor the rapid decomposition of all easily decomposable material, leaving behind mostly recalcitrant lignin as peat. Although there has been very little work in this area, there is some evidence that peat formed in warmer climates is more recalcitrant than peat formed in colder areas (Driessen and Rochimah 1976 as cited in Anderson 1983; Harden and Polak 1941 as cited in Andriesse 1988; Bridgham and Richardson 1992; Yavitt et al. 1997; Bridgham et al. 1998). Our data support this hypothesis: we also measured high concentrations of lignin in our peat samples. Changes in climate could pose a threat to the sustainability of these wetlands. Temperature is not likely to be important, even if changes did occur at this tropical location, because decomposition was limited by anaerobic conditions and poor substrate quality, neither of which would be aected by increasing temperature. However, peat accumulation could be aected by changes in precipitation, which could alter the hydrologic regime of the peatlands. For example, water levels in Yela dropped up to 0.5 m for 6 month during the 19971998 ENSO-related drought (Drexler and Ewel 2001). Water table declines of this duration and magnitude could cause a signicant loss of peat in these peatlands, suggesting that increased frequency of ENSO events could threaten the long-term integrity of these wetlands.

Management implications We found that Micronesian tropical peatlands accumulate peat not because of high plant production but rather because of slow decomposition of roots and wood. Their slow decomposition rates result from anaerobic conditions due to high water levels and poor drainage (Chimner and Ewel 2004; Allen et al. unpublished MS). Peatlands are the most widespread and numerous of all wetland types and can form under almost any climate regime if supplied with the proper hydrologic conditions. However, they are susceptible to changes in their hydrology (e.g., Chimner and Cooper 2003), and both human modication and climate change could aect hydrology in Micronesian peatlands. Farming and agroforestry are major land uses of peatlands in the tropics, with Kosrae being no exception, as about 97% of the households in Kosrae cultivate taro in wetlands (Drew et al. unpublished MS). However, we found that peat loss was likely to be slower in a Micronesian taro patch because natural hydrological conditions were maintained (Chimner and Ewel in press). The naturally high water tables in the taro patch maintained similar water chemistry, soil temperatures, decomposition rates, and soil respiration compared to an adjacent forested peatland (Yewak). In addition, any alterations to the peatlands caused by taro cultivation would probably not be permanent, because taro patches are periodically abandoned and left fallow for a few years. Thus we conclude that traditional taro cultivation is probably not a major threat to the peatlands.

Acknowledgements We thank Jason Jack, Nicole Cormier, Robert D. Hau, and Erick Waguk for assistance with eld measurements, and Brian Fry, Guy Porter, and Russell Yost for additional laboratory assistance. We are grateful to Tom Cole for help in developing volume equations. We also thank anonymous reviewers for constructive comments on the manuscript. Considerable logistic support was provided by the Kosrae Island Resource Management Authority.

References
Allen J.A., Krauss K.W., Ewel K.C., Keeland B.D. and Waguk E. unpublished MS. A tropical freshwater wetland: I. Structure, growth, and regeneration patterns. Wetlands Ecol. Manage. 13: 657669. Anderson J.A.R. 1983. The tropical peat swamps of western Malesia. In: Ecosystems of the World, 4B. Mires:swamp, Bog, Fen and Moor, Regional Studies Elsevier, Amsterdam, The Netherlands, pp. 181199. Andriesse J.P. 1988. Nature and management of tropical peat soils. FAO Soils Bulletin 59. Food and Agricultural Organization of the United Nation, Rome. Bartsch I. and Moore T.R. 1985. A preliminary investigation of primary production and decomposition in four peatlands near Scheerville, Quebec. Can. J. Bot. 63: 12411248. Bridgham S.D. and Richardson C.J. 1992. Mechanisms controlling soil respiration (CO2 and CH4) in southern peatlands. Soil Biol. Biochem. 24: 10891099.

683
Bridgham S.D., Updegra K. and Pastor J. 1998. Carbon, nitrogen, and phosphorus mineralization in northern wetlands. Ecology 79: 15451561. Brinson M.M. 1977. Decomposition and nutrient exchange of litter in an alluvial swamp forest. Ecology 58: 601609. Bums L.A. 1976. Dissolved hydrogen sulde in soil water in San Felasco Hammock and Deerhaven cypress dome. In: Odum H.T. and Ewel K.C. (eds), Cypress Wetlands for Water Management, Recycling, and Conservation. Third Annual Report to the National Science Foundation and The Rockefeller Foundation, Center for Wetlands, University of Florida, Gainesville, pp. 786802. Chamie J.P.M. and Richardson C.J. 1978. Decomposition in northern wetlands. In: Good R.E., Whigham D.F., Simpson R.L. and Jackson C.G. Jr. (eds), Freshwater Wetlands: Ecological Processes and Management Potential. Academic Press, New York. Chen R. and Twilley R.R. 1999. A simulation model of organic matter and nutrient accumulation in mangrove wetland soils. Biogeochemistry 44: 93118. Chimner R.A. 2004. Soil respiration rates in tropical peatlands of Micronesia and Hawaii. Wetlands 24: 5156. Chimner R.A. and Cooper D.J. 2003. Carbon dynamics of pristine and hydrologically modied fens in the southern Rocky Mountains. Can. J. Bot. 81: 477491. Chimner R.A., Cooper D.J. and Parton W.J. 2002. Modeling carbon accumulation in Rocky Mountain fens. Wetlands 22: 100110. Chimner R.A. and Ewel K.C. 2004. Dierences in carbon uxes between forested and cultivated Micronesian tropical peatlands. Wetlands Ecol. Manage. 12: 419427. Clymo R.S. 1983. Peat. In: Ecosystems of the World, 4B. Mires:swamp, bog, fen and moor, Regional Studies. Elsevier, Amsterdam, The Netherlands, pp. 159224. Conner W.H. and Day J.W. Jr. 1991. Leaf litter decomposition in three Louisiana freshwater forested wetland areas with dierent ooding regimes. Wetlands 11: 303312. Day F.P. Jr., Megonigal J.P. and Lee L.C. 1989. Cypress root decomposition in experimental wetland mesocosms. Wetlands 9: 263282. De la Cruz A.A. 1986. Tropical wetlands as a carbon source. Aquat. Bot. 25: 109115. Drew M.W., Ewel K.C., Naylor R.L. and Sigrah A. unpublished MS. A tropical forested wetland: III. Direct use values and ecological goods and services. Wetlands Ecol. Manage 13: 685693. Drexler J.Z. and Ewel K.C. 2001. Eect of the 19971998 ENSO-related drought on hydrology and salinity in a Micronesian wetland complex. Estuaries 24: 347356. Duever M.J., Carlson J.E., Riopelle L.A., Gunderson L.H. and Duever L.C. 1976. Ecosystem analyses at Corkscrew Swamp. In: Odum H.T. and Ewel K.C. (eds), Cypress Wetlands for Water Management, Recycling, and Conservation. Third Annual Report to the National Science Foundation and The Rockefeller Foundation, Center for Wetlands, University of Florida, Gainesville, pp. 707737. Ewel K.C., Hau R.D. and Cole T.G. 2003. Analyzing mangrove forest structure and species distribution on a Pacic island. Phytocoenologia 33: 251266. Fang C. and Moncrie J.B. 2001. The dependence of soil CO2 eux on temperature. Soil Biol. Biochem. 33: 155165. Francez A.J. and Vasander H. 1995. Peat accumulation and peat decomposition after human disturbance in French and Finnish mires. Acta Oecol. 16: 599608. Goehring H.K. and Van Soest P.J. 1970. Forage ber analyses. Apparatus, reagents, procedures, and some applications. U.S. Department of Agriculture Agricultural Handbook No. 379. U.S. Department of Agriculture, Washington, DC. Gorham E. 1991. Northern peatlands: role in the carbon cycle and probable responses to climatic warming. Ecol. Appl. 1: 182195. Gorham E., Janssens J.A. and Glaser P.H. 2003. Rates of peat accumulation during the postglacial period in 32 sites from Alaska to Newfoundland, with special emphasis on northern Minnesota. Can. J. Bot. 81: 429438. Heal O.W. and French D.D. 1974. Decomposition of organic matter in tundra. In: Holding A.J., Heal O.W., Maclean S.F. and Flanagan P.W. (eds), Soil Organisms and Decomposition in Tundra. Tundra Biome Steering Committee, Stockholm, pp. 279310. Hogg E.H., Lieers V.J. and Wein R.W. 1992. Potential carbon losses from peat proles: eects of temperature, drought cycles, and re. Ecol. Appl. 2: 298306. Hunt H.W. 1977. Simulation-model for decomposition in grasslands. Ecology 58: 469484. Immirzi C.P., Maltby E. and Clymo R.S. 1992. The global status of peatlands and their role in carbon cycling. A report for the Friends of the Earth by the Wetlands Ecosystems Research Group, Department of Geography, University of Exeter. Friends of the Earth, London. Junk W.J. 2002. Long-term environmental trends and the future of tropical wetlands. Environ. Conserv. 29: 414435. Lloyd J. and Taylor J.A. 1994. On the temperature dependence of soil respiration. Funct. Ecol. 8: 315323. Lugo A.E., Brown S. and Brinson M.M. 1988. Forested wetlands in freshwater and salt-water environments. Limnol. Oceanogr. 33: 894909. Maas A. 1996. A note on the formation of peat deposits in Indonesia. In: Maltby E., Immirzi C.P. and Saord R.J. (eds), Tropical Lowland Peatlands of Southeast Asia. Proceedings of a Workshop on Integrated Planning and Management of Tropical Lowland Peatlands. IUCN, Gland, Switzerland. MacLean C.D., Whitesell C.D., Cole T.G. and McDue K.E. 1988. Timber resources of Kosrae, Pohnpei, Truk, and Yap, Federated States of Micronesia. Resource Bulletin PSW-24, Albany, CA. Malmer N. 1986. Vegetational gradients in relation to environmental conditions in northwestern European mires. Can. J. Bot. 64: 375383. Maltby E. and Proctor M.C.F. 1996. Peatlands: their nature and role in the biosphere. In: Lappalainen E. (ed), Global Peat Resources. International Peat Society and Geological Survey of Finland, Jyska Finland. Maxwell B.D. 1982. Floristic description of native upland forests on Kosrae, Eastern Caroline Islands. Micronesica 18: 109120. Merlin M., Jano D., Raynor W., Keene T., Juvik J. and Sebastian B. 1992. Tujke en Pohnpei (Plants of Pohnpei). East-West Center, Honolulu, Hawaii, USA. Merlin M., Taulung R. and Juvik J. 1993. Sahk kap ac Kain in acn Kosrae (Plants and environments of Kosrae). East-West Center, Honolulu, Hawaii, USA.

684
Miller B. 1984. Partial inventory of Terminalia (ka) timber. Report to the State of Kosrae, Federated States of Micronesia, 22 pp. Neill C. 1992. Comparison of soil coring and ingrowth methods for measuring belowground production. Ecology 73: 1918 1921. Nessel J.K. 1978. Distribution and dynamics of organic matter and phosphorus in a sewage enriched cypress swamp. Masters thesis. University of Florida, Gainesville. Olson J.S. 1963. Energy storage and the balance of producers and decomposers in ecological systems. Ecology 44: 332 331. Parton W.J., Schimel D.S., Cole C.V. and Ojima D.S. 1987. Analysis of factors controlling soil organic matter levels in Great Plains grasslands. Soil Sci. Soc. Am. J. 51: 11731179. Raich J.W. and Potter C.S. 1995. Global patterns of carbon dioxide emissions from soils. Global Biogeochem. Cycles 9: 2336. Reader R.J. and Stewart J.M. 1972. The relationship between net primary production and accumulation for a peatland in southeastern Manitoba. Ecology 53: 10241037. Rieley J.O. and Ahmad-Shah A.A. 1996. The vegetation of tropical peat swamp forests. In: Maltby E., Immirzi C.P. and Saord R.J. (eds), Tropical Lowland Peatlands of Southeast Asia. Proceedings of a Workshop on Integrated Planning and Management of Tropical Lowland Peatlands. IUCN, Gland, Switzerland. Rieley J.O., Ahmad-Shah A.A. and Brady M.A. 1996. The extent and nature of tropical peat swamps. In: Maltby E., Immirzi C.P. and Saord R.J. (eds), Tropical Lowland Peatlands of Southeast Asia. Proceedings of a Workshop on Integrated Planning and Management of Tropical Lowland Peatlands. IUCN, Gland, Switzerland. Silver W.L. and Miya R.K. 2001. Global patterns in root decomposition: comparisons of climate and litter quality eects. Oecologia 129: 407419. Soil Survey of Island of Kosrae, Federated States of Micronesia. 1983. USDA Department of Agriculture. Soil Conservation Service. SPSS. 2000. SYSTAT 10.0 for Windows. SPSS, Chicago, Illinois, USA. Suzuki S., Ishida T., Nagano T. and Waijaroen S. 1999. Inuences of deforestation on carbon balance in a natural tropical peat swamp forest in Thailand. Environ. Control Biol. 37: 115128. Thormann M.N. and Bayley S.E. 1997. Decomposition along a moderate-rich fen-marsh peatland gradient in boreal Alberta, Canada. Wetlands 17: 123137. Thormann M.N., Bayley S.E. and Currah R.S. 2001. Comparison of decomposition of belowground and aboveground plant litters in peatlands of boreal Alberta, Canada. Can. J. Bot. 79: 922. Trettin C.C., Jurgensen M.F., Gale M.R. and McLaughlin J.W. 1995. Soil carbon in northern wetlands: impacts of silvicultural practices. In: Carbon Forms and Functions in Forest Soils. Soil Science Society of America, Madison, WI. Trumbore S.E., Bubier J.L., Harden J.W. and Crill P.M. 1999. Carbon cycling in boreal wetlands: a comparison of three approaches. J. Geophys. Res.Atmos. 104(D22): 27673 27682. Tupacz E.G. and Day F.P. 1990. Decomposition of roots in a seasonally ooded swamp ecosystem. Aquat. Bot. 37: 199 214. Verhoeven J.T.A. 1986. Nutrient dynamics in minerotrophic peat mires. Aquat. Bot. 25: 117137. Vijarnsorn P. 1996. Peatlands in Southeast Asia: a regional perspective. In: Maltby E., Immirzi C.P. and Saord R.J. (eds), Tropical Lowland Peatlands of Southeast Asia. Proceedings of a Workshop on Integrated Planning and Management of Tropical Lowland Peatlands. IUCN, Gland, Switzerland. Weltzin J.F., Pastor J., Harth C., Bridgham S.D., Updegra K. and Chapin C.T. 2000. Response of bog and fen plant communities to warming and water table manipulations. Ecology 81: 34643478. Wieder R.K. and Lang G.E. 1982. A critique of the analytical methods used in examining decomposition data obtained from litter bags. Ecology 63: 16361642. Yavitt J.B., Williams C.J. and Wieder R.K. 1997. Production of methane and carbon dioxide in peatland ecosystems across North America: eects of temperature, aeration, and organic chemistry of peat. Geomicrobiology 14: 299316.